1. Field of the Invention
The present invention relates to a method of and a system for charged particle beam exposure on semiconductor wafers, masks and so on having a coating of resist.
2. Description of the Related Art
As the elements in semiconductor integrated circuits become ever finer, the use of charged particle beam exposure systems will become common, and in such systems, electrons may be used as charged particles. With these systems, it is possible to produce very fine elements, down to 0.05 .mu.m or smaller with alignment accuracy of 0.02 .mu.m or better.
FIG. 1 shows the structure of parts of a charged particle beam exposure system in the prior art.
The object to be exposed here is a semiconductor wafer 10 that is mounted on a mobile stage 11. The wafer 10 is coated with a film of resist and a charged particle beam, in this instance, an electron beam EB, is radiated onto the film to perform exposure.
Above the wafer 10, a blanking aperture array 13 is provided over an aperture 12. As shown in FIG. 2, the blanking aperture array 13 is provided with a plurality of openings 15A to 15C formed in grid pattern in a thin substrate and is also provided with common electrodes 16A to 16C and blanking electrodes 17A to 17C which project downward and which are formed at the edges of the openings 15A to 15C respectively. The openings 15A to 15C may be, for example, squares whose sides are 25 .mu.m, and the number of these openings may be 16.times.64. An electron beam EB, which is made to run approximately parallel by an electromagnetic lens 19 and which has a nearly consistent current density, is radiated on to the blanking aperture array 13. The point of irradiation on the wafer 10 of a portion of the electron beam that has passed through one opening 15 may be, for example, an approximate square with sides of 0.08 .mu.m.
By setting the voltage between the common electrodes 16A to 16C and the blanking electrodes 17A to 17C to 0 or Vs, the electron beam that passes through the openings 15A to 15C is either radiated onto the wafer 10 after passing through the aperture 12, shown in FIG. 1, or it is blocked by the aperture 12. As a result, by setting the common electrodes 16A to 16C at 0V and also by supplying a voltage pattern that corresponds to the pattern to be exposed to all the blanking electrodes 17A to 17C from a blanking control circuit 14, a very fine pattern can be exposed onto the semiconductor wafer 10.
However, since exposure is performed by the charged particle beam scanning over the semiconductor wafer or a mask, the exposure takes longer than in photo exposure. Because of this, it is necessary to improve the throughput of the charged particle beam exposure system. Generally speaking, improving the throughput also entails a loss of exposure positional accuracy, which results in a reduction in processing accuracy. Thus, an improvement in exposure positional accuracy is required together with an improvement in throughput.
Scanning the electron beam EB on the wafer 10 is performed by the mobile stage 11, and a main deflector 20 and a sub deflector 30 that are positioned over the mobile stage 11. Normally, the main deflector 20 is an electromagnetic type and the sub deflector 30 is an electrostatic type.
The ranges of movement from the greatest to the least whereby scanning can be performed are provided by: the mobile stage 11, the main deflector 20 and the sub deflector 30, in that order. However, the order of scanning speed is the reverse, with the sub deflector 30 first, the main deflector 20 next, and the mobile stage 11 last. By taking advantage of these scanning characteristics, scanning with the electron beam EB on the wafer 10 is performed, for example, as shown in FIG. 3.
That is, the electron beam EB is scanned by the sub deflector 30 within a subfield F in direction A. Every time the scanning within one subfield F is completed, step scanning is performed by the main deflector 20 by the width of the subfield F in the direction B of the stripe ST (in the direction which runs at a right angle to the direction A). Also, with the mobile stage 11, scanning in direction C which is perpendicular to direction B is continuously performed. The length of the stripe ST may be 2 mm, with one side of the subfield F being 100 .mu.m, for instance.
(1) Reduction of throughput in an exposure system imposed by the setting time.
In FIG. 1, the main deflection data DM are converted to analog by a D/A converter 21 and then this data DM is amplified at an amplifying circuit 22 to be supplied to the main deflector 20 as a drive current I. Likewise, the sub deflection data DS is converted to analog by D/A converter 31 and then this data DS is amplified at an amplifying circuit 32 to be supplied to the sub deflector 30 as a drive voltage V.
When scanning with the electron beam EB in direction A in FIG. 3 with the sub deflector 30, the drive voltage V changes in a small step for every pulse of the electron beam EB and when swinging back from the trailing end to the leading end of a scan in direction A, the drive voltage V changes in a large step, in correspondence with the dotted lines. FIG. 4A shows the waveform of the drive voltage V when it changes in a small step and FIG. 4B shows the waveform of the drive voltage V when it changes in a large step.
Due to the frequency characteristics of the amplifying circuit 32 and the capacity of the sub deflector 30, the setting time of the drive voltage V cannot be set to 0 when the input of the amplifying circuit 32 changes in steps. During this setting time, exposure cannot be performed because the deflecting error is significant. Therefore, it is wait time and it is a cause of reduced throughput of the charged particle beam exposure system.
Since the number of changes is larger when the changes are in small steps than when they are in in large steps, in the prior art, the characteristics of the amplifying circuit 32 are adjusted in such a manner that the setting time .DELTA.t1 for the small step change is at a minimum. Because of this, if the setting time .DELTA.t1 is set at 100 ns, for example, the setting time .DELTA.t2 for the large step change is a relatively large value, i.e., 500 ns. This prevents any improvement in throughput in the charged particle beam exposure system. The aforementioned adjustment is performed by changing the electrostatic capacity of the capacitor that is included in the amplifying circuit 32, through which high frequency components pass.
On the other hand, there is a significant transient waveform included in the outputs of both of the D/A converters 21 and 31 in FIG. 1, and the their setting time also presents a cause for reduction in throughput of the exposure system.
In the D/A converters, as shown in FIG. 5A, for instance, the output ends of constant current sources 210 to 213, the input ends of which are commonly connected to a source wire VCC, are connected to the input end of a operational amplifier 218 via analog switches 214 to 217 respectively, and the resistor 219 is connected between the input end and the output end of the operational amplifier 218. The constant currents I, 2I, 4I and 8I are output from the constant current sources 210 to 213 respectively.
When simultaneously turning ON the analog switches 214 to 217, from the state in which the analog switches 214 to 217 are all OFF, because of differences in operating speed among the analog switches 214 to 217 and the like, the analog switches 214 to 217 are not turned ON perfectly simultaneously. As a result, a transient waveform (glitch waveform) as shown in FIG. 5B is included in the output of the operational amplifier 218. Because of its origin, the magnitude of this glitch waveform depends upon the values at the input of the D/A converters before and after the change.
(2) Reduction in throughput of an exposure system due to a reduction in the deflection area of an electrostatic deflector
FIG. 6A is a perspective view of an electrostatic deflector in the prior art and FIG. 6B is a cross section through the line 6B--6B in FIG. 6A.
Four pairs of electrostatic deflecting electrodes a1 to a4 and b1 to b4 that face opposite each other are formed by plating a conductive material onto the inner surface of a cylindrical member 301 which is formed in one piece by injection. A wire 302 for applying a voltage from the outside is connected to the end of each electrode. Reference number 303 indicates a portion where the plating is cut and the other corresponding portions are identical.
With the four pairs of electrodes, a consistent electrical field is formed over a wider region compared with an arrangement with only two pairs of electrodes that are positioned at a right angle to each other to cause deflection in direction X and in direction Y. In each pair of electrodes, one is set at 0V, or, potentials whose absolute values are equal to each other but whose signs are opposite are applied from a complementary output type amplifier.
FIG. 7 shows the path that is obtained when an electron beam EB is deflected in the direction indicated with the arrow A1 with a one-stage electrostatic deflector 30A within an electromagnetic lens (not shown). Since the magnetic field of the electromagnetic lens causes the electron beam EB to converge, the electron beam EB running along the z-axis is deflected as indicated with the arrow. This path turns spirally as indicated with the dotted line when viewed in the x-y plane which contains the x-axis and the y-axis, and which lies at a right angle to the z-axis, restricting the deflection area.
With a reduced sub deflection area, the number of times the main deflector is operated increases and, on the other hand, since the setting time for a deflection imparted by the main deflector is longer than for one imparted by the sub deflector, the throughput of the electron beam exposure system is reduced.
This problem becomes more pronounced when an electrostatic deflector 30 is positioned within an immersion lens 40, which is constituted with the lens portions 40A and 40B provided above and below the wafer 10. The immersion lens 40 achieves low aberration and high resolution by forming a strong magnetic field on the wafer 10 to converge the electron beam EB, making more fine processing possible.
The following methods may be employed in order to increase the sub deflection area:
(a) Increase the deflection voltage. PA1 (b) Reduce the distance between electrostatic deflecting electrodes that face opposite each other. PA1 (c) Lengthen electrostatic deflecting electrodes.
However, since an amplifier capable of high voltage output has inferior response characteristics, with method (a), the throughput of an electron beam exposure system is reduced. With method (b) or (c), the processing accuracy on the inner surface of the cylindrical member 301 is reduced and, consequently, the deflection accuracy is also reduced. As the path of the electron beam EB within the immersion lens 40 rotates spirally, only a very small increase in deflection area is achieved with method (c).
(3) Reduction in throughput of an exposure system caused by a long shift time between subfields required by an electromagnetic deflector
In order to utilize an electron beam exposure system for LSI mass production, the total of exposure time per wafer must be equal to or less than 3 minutes.
As shown in FIG. 11, for instance, electromagnetic deflectors 20A to 20D are positioned in four stages in the direction of the optical axis and their coils are connected in series with the number of turns at, for example, 80 and the inductance at approximately 50 .mu.H. Because of this, the response is low and the shift time between the subfields F in FIG. 3 is, for example, 50 .mu.sec. The number of subfields in a wafer 8" in diameter is typically 2.4.times.10.sup.6 making the total length of shift time 120 sec. This would make it impossible to utilize an electron beam exposure system for LSI mass production.
In order to solve this problem, a method has been disclosed (First Publication No. 62-277724 of Japanese Patent Application, U.S. Pat. No. 4,853,870) in which the difference V1X-V2X between the output V1X of the D/A converter 21 and the monitor output V2X of the amplifying circuit 22 in FIGS. 1 and 9 is detected, to be added to the input of the differential amplifying circuit 32.
However, while the deflection area within subfields is -50 to 50 .mu.m, a voltage capable of deflecting by approximately 100 .mu.m must be applied to the sub deflector 30 at the rise of the output V1X of the D/A converter 21 during a shift between subfields and it is necessary, therefore, to approximately double the amplification factor of the amplifying circuit 30. This reduces the SN ratio, which in turn causes a problem that the exposure positional accuracy when scanning within subfields with the sub deflector 30 is reduced.
(4) Reduction in throughput of an exposure system due to long alignment time of patterns between layers.
During the manufacturing process of a semiconductor integrated circuit, it is necessary to mode a circuit pattern with aligning several tens of layers on a semiconductor wafer. The minimum pattern size of a 256 Mbit DRAM is 0.25 .mu.m, for example, and the tolerance for layer alignment is equal to or less than 1/8 of that, i.e., 0.031 .mu.m, which means that a very high degree of alignment accuracy is required.
As shown in FIG. 10, in an electron beam exposure process, chip areas 10i, 1=1 to n are formed on the semiconductor wafer 10 and around each chip area 10i, aligning marks Ai, Bi, Ci and Di for layer alignment are formed. When exposure is performed with a stepping projection aligner, the global alignment method, in which, of n number of chip areas 10i, the positions of the aligning marks Ai for 3 chips areas 10i are detected, for instance, and based upon the relationships with the corresponding positions in design, a relational expression that expresses the relationship between the design position of a given aligning mark Ai and its actual position and the semiconductor wafer 10 is determined to be used for alignment.
However, since a high degree of alignment accuracy is required in electron beam exposure, as mentioned earlier, the di-by-di alignment method is employed, wherein the aligning marks Ai, Bi, Ci and Di for each chip area 10i are detected for alignment. For example, when 50 chip areas are formed on a semiconductor wafer 10 with a 6-inch diameter, while electron beam exposure for one layer takes 2 minutes, it takes 7 minutes to detect all the aligning marks, resulting in low throughput of electron beam exposure.
(5) Reduction in the exposure pattern accuracy due to astigmatism of an electromagnetic lens
As shown in FIG. 11, electromagnetic main deflectors 20A1 to 20A4 and an astigmatism correction coil 41, for deflecting an electron beam EB, are provided on the inside of an electromagnetic lens 40A. An electron beam EB0 diverges from the point of intersection of the optical axis 42 and the lateral cross sectional plane 43 of the electron beam EB when the deflection quantity is 0, and is converged onto the wafer 10 by the electromagnetic lens 40A as indicated with the solid line.
FIG. 12A shows a situation in which the astigmatism correction coil 41 is not operated. Because of the astigmatism of the electromagnetic lens 40A, the cross-over point CX of the X-Z cross section plane 44X and the cross-over point CY of the Y-Z cross section plane 44Y of the electron beam are offset, and the electron beam spot on the wafer becomes rectangular as indicated by the shaded area. If there is no astigmatism present, the electron beam spot 46 is square. The astigmatism correction coil 41 generates a magnetic field which expands or contracts the shape of the lateral cross sectional plane of an electron beam to correct the astigmatism.
The X-Y cross section plane 44X, the Y-Z cross section plane 44Y of the electron beam and its spot 45 on the wafer when the position PZ of the astigmatism correction coil 41 on the optical axis is set at PZ&gt;CX, PZ=CX, CY&lt;PZ&lt;CX, PZ=CY and PZ&lt;CY for the astigmatism at the Y-Z cross section 44Y in FIG. 12A are shown in FIG. 12B to 12F respectively. Reference numbers 471 to 474 indicate the directions, i.e., direction X and direction Y of the electromagnetic force applied to the electron beam EB by the magnetic field generated by the astigmatism correction coil 41. For the sake of simplification, it is assumed that the cross section of the electron beam instantaneously changes when it is at a position in direction Z of the astigmatism correction coil 41.
Since, on the optical axis 42, the electromagnetic force generated by the astigmatism correction coil 41 cancels itself to 0, in FIG. 12C there is no change in the X-Z cross section 44X of the electron beam and only the Y-Z cross section 44Y is enlarged at a position in direction Z of the astigmatism correction coil 41. Consequently, by adjusting the electric current supplied to the astigmatism correction coil 41, the electron beam spot 45 can be aligned with the electron beam spot 46 obtained when there is no astigmatism.
However, cross-over points vary depending upon the position within the electromagnetic lens 40A, i.e., depending upon the deflection position of the electron beam EB and, therefore, the astigmatism correction coil 41 cannot be position aligned with the cross-over point CX. It is the same when there is an astigmatism generated at the X-Z cross section 44X.
Actually, it is verified that the shape of the cross section of the electron beam EB distorts at the corners of the deflection area. For example, compared to the regular shape, the shape of the cross section of the electron beam becomes flattened at the upper right corner, and it becomes oblong at the lower left corner. When the cross section of the electron beam EB is a 3 .mu.m.times.3 .mu.m square, for instance, the quantity of change in the shape of the cross section is equal to or less than 0.1 .mu.m for each side. In order to meet with the requirement of more fine patterns, it is necessary to raise the accuracy of astigmatism correction to a level higher than that achieved currently.